mirror of
				https://github.com/yt-dlp/yt-dlp.git
				synced 2025-11-04 08:30:46 +00:00 
			
		
		
		
	
		
			
				
	
	
		
			80 lines
		
	
	
		
			2.8 KiB
		
	
	
	
		
			Python
		
	
	
	
	
	
			
		
		
	
	
			80 lines
		
	
	
		
			2.8 KiB
		
	
	
	
		
			Python
		
	
	
	
	
	
from __future__ import unicode_literals
 | 
						|
 | 
						|
import re
 | 
						|
import datetime
 | 
						|
 | 
						|
from .common import InfoExtractor
 | 
						|
 | 
						|
 | 
						|
class VubeIE(InfoExtractor):
 | 
						|
    IE_NAME = 'vube'
 | 
						|
    IE_DESC = 'Vube.com'
 | 
						|
    _VALID_URL = r'http://vube\.com/[^/]+/(?P<id>[\da-zA-Z]{10})'
 | 
						|
 | 
						|
    _TEST = {
 | 
						|
        'url': 'http://vube.com/Chiara+Grispo+Video+Channel/YL2qNPkqon',
 | 
						|
        'md5': 'f81dcf6d0448e3291f54380181695821',
 | 
						|
        'info_dict': {
 | 
						|
            'id': 'YL2qNPkqon',
 | 
						|
            'ext': 'mp4',
 | 
						|
            'title': 'Chiara Grispo - Price Tag by Jessie J',
 | 
						|
            'description': 'md5:8ea652a1f36818352428cb5134933313',
 | 
						|
            'thumbnail': 'http://frame.thestaticvube.com/snap/228x128/102e7e63057-5ebc-4f5c-4065-6ce4ebde131f.jpg',
 | 
						|
            'uploader': 'Chiara.Grispo',
 | 
						|
            'uploader_id': '1u3hX0znhP',
 | 
						|
            'upload_date': '20140103',
 | 
						|
            'duration': 170.56
 | 
						|
        }
 | 
						|
    }
 | 
						|
 | 
						|
    def _real_extract(self, url):
 | 
						|
        mobj = re.match(self._VALID_URL, url)
 | 
						|
        video_id = mobj.group('id')
 | 
						|
 | 
						|
        video = self._download_json('http://vube.com/api/v2/video/%s' % video_id,
 | 
						|
            video_id, 'Downloading video JSON')
 | 
						|
 | 
						|
        public_id = video['public_id']
 | 
						|
 | 
						|
        formats = [{'url': 'http://video.thestaticvube.com/video/%s/%s.mp4' % (fmt['media_resolution_id'], public_id),
 | 
						|
                   'height': int(fmt['height']),
 | 
						|
                   'abr': int(fmt['audio_bitrate']),
 | 
						|
                   'vbr': int(fmt['video_bitrate']),
 | 
						|
                   'format_id': fmt['media_resolution_id']
 | 
						|
                   } for fmt in video['mtm'] if fmt['transcoding_status'] == 'processed']
 | 
						|
 | 
						|
        self._sort_formats(formats)
 | 
						|
 | 
						|
        title = video['title']
 | 
						|
        description = video.get('description')
 | 
						|
        thumbnail = video['thumbnail_src']
 | 
						|
        if thumbnail.startswith('//'):
 | 
						|
            thumbnail = 'http:' + thumbnail
 | 
						|
        uploader = video['user_alias']
 | 
						|
        uploader_id = video['user_url_id']
 | 
						|
        upload_date = datetime.datetime.fromtimestamp(int(video['upload_time'])).strftime('%Y%m%d')
 | 
						|
        duration = video['duration']
 | 
						|
        view_count = video['raw_view_count']
 | 
						|
        like_count = video['total_likes']
 | 
						|
        dislike_count= video['total_hates']
 | 
						|
 | 
						|
        comment = self._download_json('http://vube.com/api/video/%s/comment' % video_id,
 | 
						|
            video_id, 'Downloading video comment JSON')
 | 
						|
 | 
						|
        comment_count = comment['total']
 | 
						|
 | 
						|
        return {
 | 
						|
            'id': video_id,
 | 
						|
            'formats': formats,
 | 
						|
            'title': title,
 | 
						|
            'description': description,
 | 
						|
            'thumbnail': thumbnail,
 | 
						|
            'uploader': uploader,
 | 
						|
            'uploader_id': uploader_id,
 | 
						|
            'upload_date': upload_date,
 | 
						|
            'duration': duration,
 | 
						|
            'view_count': view_count,
 | 
						|
            'like_count': like_count,
 | 
						|
            'dislike_count': dislike_count,
 | 
						|
            'comment_count': comment_count,
 | 
						|
        } |